In this section we consider the genetic evidence for reproductive isolation between populations or groups of populations. Allele-frequency differentiation among populations and differences in levels of gene diversity constitute the bulk of this evidence. Populations of chum salmon have been examined for genetic variability throughout most of their geographical range around the North Pacific Rim. Most of these studies were made with protein electrophoresis, but recent studies have also used the analysis of mitochondrial DNA (mtDNA) and VNTR (variable number of tandom repeats) microsatellite DNA.
One approach to detecting reproductive isolation is to compare frequencies of protein variants (allozymes) among samples with a contingency-table analysis and the chi-square statistic or the G-statistic (log likelihood ratio statistic). The finding of significant frequency differences between groups can be taken as evidence of reproductive isolation. Another approach to identifying reproductively isolated groups is to analyze genetic distances between samples with such clustering methods as the unweighted pair groups method with averages (UPGMA) (Sneath and Sokal 1973) or multidimensional scaling (MDS) (Kruskal 1964). The UPGMA method is appropriate for analyzing populations of species such as salmon that largely have hierarchical genetic population structure, with large geographical groups divided into subgroups. However, when genetic variability varies continuously, as in a cline, or when geographically intermediate populations are also genetically intermediate between groups, MDS more accurately depicts relationships among samples than does agglomerative clustering such as the UPGMA (Lessa 1990). MDS is a nonmetric ordination of samples in two or three dimensions that represents dissimilarity among samples. Principle component analysis (PCA) of allelic frequencies can also be used to examine genetic relationships among populations. The results of PCA analysis are usually similar to the results of MDS analysis for a set of data.
Several genetic distance measures (e.g., Cavalli-Sforza and Edwards 1967; Rogers 1972; Nei 1972, 1978) have been used to study the population genetic structure of chum salmon, as well as other salmonids. It is unclear, however, which measure is best. An attractive feature of Rogers' and Cavalli-Sforza and Edwards' distances is that they are metrics and satisfy the triangle inequality: Given three populations (A, B, C), the sum of the distances between A and B and between B and C is greater than or equal to the distance between A and C. Nei's (1972, 1978) distances sometimes distort genetic distance so that the triangle inequality is not met. On the other hand, Cavalli-Sforza and Edwards' and Rogers' genetic distances do not use a correction for sample size; thus distances are biased upward, especially for small sample sizes (Nei 1978). In contrast, Nei's (1978) distance is unbiased.
When sample sizes used to estimate allelic frequencies are 50 individuals or more, the difference between Nei's genetic distance (Nei 1972) and Nei's unbiased genetic distance (Nei 1978) is small. Sample sizes much less than 50 individuals may inflate estimates of genetic distance between populations, because of the increased error in estimating allele frequencies. An unbiased statistic is therefore more desirable. However, when genetic distances between populations are also small, as they often are between populations of salmon, low but significant levels of genetic differentiation may not be detected by an unbiased distance measure because sample-size corrections may reduce estimates of genetic distance to zero.
Another consideration is that Nei's (1972, 1978) and Rogers' (1972a) distance measures can be influenced by different levels of heterozygosity between populations, whereas Cavalli-Sforza and Edwards' (1967) measure cannot. Discussions of these and other features of genetic distances appear in Nei (1978), Hillis et al. (1996), and Rogers (1991). Most of this discussion has focused on the merits of the various measures for phylogenetic reconstruction among species and higher taxa. No one has quantitatively evaluated the performances of these distances in assessing genetic differentiation among populations of species like salmon, which typically show small genetic distances between conspecific populations.
Since it is unclear which distance measure is "best" in any given application, we analyzed each set of data with Nei's unbiased (1978), Rogers' (1972a), and Cavalli-Sforza and Edwards' (1967) genetic distances to identify results that may not be robust to the distance measure used. In most cases, the different genetic distance measures yielded results that were highly correlated. For simplicity we report only results based on Cavalli-Sforza and Edwards' distance measure. This measure ranges from 0.0 (identity) to 1.0 (complete dissimilarity).
A gene diversity analysis (Nei 1973, Chakvaborty 1980) was used to apportion allele-frequency variability into its geographic or behavioral components for a regional data set. Most genetic variability in salmonids occurs as genotypic difference among individuals within a population. A smaller proportion is generally due to genetic differentiation between spawning populations in rivers and streams that have been defined by geography or run timing. These statistics facilitate comparisons among regions and may detect regional differences in gene flow or the effects of hatchery strays on genetic population structure.
Several life-history characteristics in chum salmon may influence the population genetic structure. Fish return to natal streams at 3-5 years of age and spawn as early as June in northern areas and as late as March in southern areas. Overlapping generations of fish at spawning localities tend to increase the effective size of a population in a river, decreasing the effects of random-drift genetic variability within and between populations. Since chum salmon generally do not migrate far into a river to spawn, less genetic differentiation might be expected relative to other species of salmon which have more geographically isolated populations scattered throughout a river system.
A major east-west genetic subdivision of Pacific Rim chum salmon populations has been detected in several studies. In an ocean-wide study, Okazaki (1983) found a major genetic discontinuity between Asian and North American chum salmon; however, because his study included only a few samples, he was not able to resolve the geographic boundaries of these groups. Seeb et al. (1995) combined data for 200 populations around the Pacific Rim from studies of Kondzela et al. (1994), Phelps et al. (1994), Seeb et al. (1995), Wilmot et al. (1994), and Winans et al. (1994). In this data set, genetically homogeneous populations were pooled into regional groups, and some alleles occurring at low frequencies were pooled with alleles with similar electrophoretic mobilities. Analyses of Cavalli-Sforza and Edwards' (1967) genetic chord distances between samples resolved a large geographic subdivision among Pacific Rim chum salmon populations, with the break occurring to the east (Wilmot et al. 1994, Seeb et al. 1995) of the break suggested by Okazaki (1983). The results of Seeb et al. (1995) indicated that northwestern Alaskan populations were genetically more similar to Asian populations than they were to populations on the Alaska Peninsula and those in central and southeastern Alaska.
The analysis of mini-satellite nuclear DNA genes, however, showed a different geographical pattern of genetic subdivision. Taylor et al. (1994) examined variability for two putative VNTR loci in 1,211 fish from 39 localities extending from Honshu Island, Japan to southern British Columbia, including localities in the Russian Federation, the Yukon River, and southeastern Alaska. A UPGMA cluster analysis of a generalized genetic distance based on band frequencies pooled into 44 size classes showed three hierarchical groups in which the greatest subdivision was between Japanese samples and the remaining samples, and the next lower subdivision was between Russia/Yukon River samples and southeastern Alaska/British Columbia samples. The differences in population structure suggested by allozymes and by nuclear mini-satellite DNA may be due to the analysis of only two linked mini-satellite loci.
The analyses of nucleotide sequence variability in mitochondrial DNA by Park et al. (1993) showed a pattern of regional differentiation similar to that indicated by the analysis of mini-satellite DNA by Taylor et al. (1994). Park et al. (1993) found a lack of sequence variability in the control region (D-loop) of mtDNA in 798 fish collected from 42 localities extending across the Pacific Rim from Japan to Washington. However, they found restriction fragment variability in polymerase chain reaction (PCR) amplified mtDNA fragments of the NADH dehydrogenase gene that showed strong frequency differences across the Pacific Rim. The largest frequency differences were between Japan and all other regions including the Russian Federation, Alaska/Yukon, and British Columbia/Washington. The mean frequency of this variant in Japan was 0.80, whereas its frequency in the other regions was 0.13 or less.
On the Asian side of the North Pacific, Japanese and Russian populations constitute genetically discrete geographic units. Winans et al. (1994) compared 17 Japanese populations on Honshu and Hokkaido Islands with 12 Russian populations extending from the Okhotsk Sea in the south to the Anadyr River in the north. Cluster analysis of Nei's unbiased genetic distances (Nei 1978) between samples and a principal component analysis (Winans et al. 1994) showed that the Japanese and Russian chum were genetically distinct from one another, with an average genetic distance of 0.006 for a sample of 62 isozyme loci. This genetic distance is near the low end of the range typically found for conspecific populations (Thorpe 1982).
Ninety percent of chum salmon produced in Japan are raised in about 300 hatcheries (Kaeriyama 1989) located on most of the rivers that historically had runs of chum salmon. In a study of 43 populations on Hokkaido and Honshu Islands, Okazaki (1982a,b) reported genetic subdivision between populations on the two islands and, to a lesser degree, between areas on each island. We reexamined the published allelic frequencies with genetic distances and cluster analyses and found no evidence of regional or local genetic groupings among Japanese populations of chum salmon. Our gene-diversity analysis of these data indicated a small amount of genetic differentiation among populations. Of the total genetic variability, 96% was contained on average within populations, 3.1% was due to differences among populations within four regions, 0.4% was due to differences between islands, and 0.5% was due to the average difference between east and west coast populations on each island.
Winans et al. (1994) examined an overlapping set of 17 populations on Hokkaido and the northern part of Honshu Island for geographic variability at 26 polymorphic loci. Their UPGMA clustering of Nei's genetic distance between populations for these data indicated that the greatest amount of geographic differentiation was between east and west coast populations and not between Hokkaido and Honshu Islands, as suggested by Okazaki (1982a,b).
Kijima and Fujio (1982) examined genetic relationships among populations of chum salmon in Japan based on allele-frequency variability at six loci. They searched for correlations between genetic distances and geographical distances between samples with four different analyses of population groups delimited by different potential pathways of migration around Hokkaido and Honshu. In the genetic-distance/geographical-distance comparisons, the magnitude of genetic distances between pairs of populations increased for a separation of up to 600 km between the pairs of samples. Pairs of samples separated by more than 600 km showed no correlation between genetic distance and geographic distance. This was interpreted to indicate that gene exchange between populations was largely limited to distances less than 600 km and that populations separated by more than 600 km were genetically independent of one another.
In contrast to Japanese chum salmon populations, Russian populations appear to be subdivided into two genetically distinct groups with a few genetic outliers. In a genetic study of chum salmon collected at river mouths, Viktorovskii et al. (1986) found two large groups, one located on the eastern Kamchatka Peninsula and the other on the western side of the Peninsula and around the Sea of Okhotsk. The southernmost populations on Sakhalin Island and around the Bay of Amur were genetic outliers. The results of Winans et al. (1994) generally confirmed the existence of two large Russian populations. In that study, samples of chum salmon were collected from 17 localities in the Russian Federation, extending from the Ola River on the Okhotsk Sea in the south around the Kamchatka Peninsula to the Anadyr River in the north, and were examined for variability at 35 loci. A UPGMA analysis of Nei's unbiased genetic distance produced four clusters that were geographically poorly resolved. Nonetheless, populations around the Okhotsk Sea (mainland and western Kamchatka Peninsula) tended to be genetically distinct from those of the eastern Kamchatka Peninsula.
Northwestern Alaska/Yukon--Chum salmon of northwestern Alaska and the Yukon River exhibit diverse life-history patterns and run times. Some chum salmon, for example, migrate 2,000 miles up the Yukon River to spawn, whereas others spawn in the lower sections of the Yukon River basin. In a study of Alaskan and Canadian chum salmon spawning in the Yukon River and its tributaries, Beacham et al. (1988) examined 7 polymorphic loci in 10 populations, including those spawning several hundred miles from the mouth of the Yukon River. They found that fall-run fish of the upper Yukon River were genetically distinct from summer-run fish in the lower Yukon River in northwestern Alaska. Other genetic studies (Seeb et al. 1995, Wilmot et al. 1994) also confirmed major genetic differences between upper Yukon fall-run chum salmon and summer-run chum salmon in the lower river. The large genetic differences between these two groups of populations appear to reflect large differences in run times and large geographic distances between spawning areas.
Wilmot et al. (1994) examined 24 polymorphic loci (plus 30 invariant loci) in 30 river populations around Bristol Bay, in the Yukon River, and on the Alaska Peninsula. Three samples from the Russian Federation were also included in this study. In a UPGMA and neighbor-joining cluster analysis of chord genetic distances, they found three major clusters of samples: 1) Russia-Alaska Peninsula samples, 2) Bristol Bay samples, and 3) upper and lower Yukon River samples. A hierarchical gene-diversity analysis of allele-frequency variability indicated that 95.4% of gene diversity was contained, on average, within populations, 1.4% was due to allele-frequency differences among populations within regions, 0.5% to differences between fall- and summer-spawning populations in the Yukon River, and 2.7% to differences among areas within years.
Southeastern Alaska/British Columbia--Kondzela et al. (1994) examined 42 variable loci (plus 4 monomorphic loci) in 61 samples of chum salmon from populations in southeastern Alaska and northern British Columbia. A neighbor-joining cluster analysis of chord distances (based on 40 loci) showed four more or less distinct clusters of populations: 1) central southeastern Alaska, 2) southern southeastern Alaska and north and central British Columbia, 3) Prince of Wales Islands, and 4) Queen Charlotte Islands. A gene-diversity analysis of these data indicated that 97% of the genetic diversity occurred as differences between individuals within populations, 1.3% as allele-frequency differences among populations within regions, and 1.4% as frequency differences among regions.
British Columbia--Beacham et al. (1987) studied allozyme variability for 9 polymorphic loci in 83 chum salmon populations in central and southern British Columbia. A UPGMA tree of Nei's unbiased genetic distance indicated five more or less distinct clusters of samples: 1) Queen Charlotte Island samples, 2) north and central coast British Columbia samples, 3) east coast of Vancouver Island and south coast of British Columbia samples, 4) west coast of Vancouver Island samples, and 5) Fraser River samples. This study did not include samples outside British Columbia, thus the relationships of these groups to chum salmon populations to the north in Alaska or to the south in Washington are unclear. An analysis of geographic variability with the fixation index (FST) indicated that on average 97.7% of the total gene diversity was contained within populations, and 2.3% was due to allele-frequency variability differences among populations.
Washington/British Columbia--Phelps et al. (1994) examined genetic variability at 39 polymorphic loci in 153 samples from 105 locations in southern British Columbia, Washington, and Oregon (Table 11). Five of 30 spawning localities examined for interannual variability showed significant (P < 0.05) allele-frequency differences between or among years. Genetic marking or hatchery transplantations could explain the temporal variability in three of these populations, and temporal variability at two localities with small population sizes was probably due to random genetic drift. Allelic frequencies were pooled over years for those populations sampled in more than 1 year, except for the Samish Hatchery which had a recent history of transfers that apparently have produced significant allele-frequency shifts among years.
A UPGMA tree of chord genetic distances consisted of clusters with samples from the same geographical region or run-time. The most important result of this analysis was that summer-run chum salmon in Hood Canal and the Strait of Juan de Fuca were distinct from fall-run chum salmon in the same areas and from other fall- and summer-run populations. This genetic distinction also appeared in the multidimensional scaling analysis of the chord distances. Unlike Hood Canal and Strait of Juan de Fuca summer-run chum salmon, the summer-run chum salmon of southern Puget Sound clustered with fall-run chum salmon of the same geographical region.
Another important result was that among fall-run chum salmon, geographically close populations were generally more similar genetically to one another than to widely separated populations. One exception was fall-run chum salmon in the Strait of Juan de Fuca and coastal Washington and Oregon, which were more similar to Georgia Strait and west-coast Vancouver Island populations than to populations in Puget Sound. Samples from two southern Puget Sound winter-run populations were included in the study, and these two samples were genetic outliers that were most closely related to samples of fall-run Hood Canal and northern Puget Sound populations (Fig. 12).
Phelps24 (and Phelps 1995) added allele-frequency data for an additional 16 chum salmon populations to the data of Phelps et al. (1994) and resolved three reasonably distinct clusters of samples: 1) summer-run chum salmon of Hood Canal and Strait of Juan de Fuca, 2) Puget Sound fall-run and southern Puget Sound winter- and summer-run chum salmon, and 3) Strait of Juan de Fuca, coastal Washington, and Oregon fall-run chum salmon. Samples from British Columbia were not included in this second analysis, but the previous results of Phelps et al. (1994) indicated that group 3 was most closely related to Fraser River and Georgia Strait chum salmon populations in British Columbia.
A gene-diversity analysis of the 105 populations sampled by Phelps et al. (1994) was typical of that for other regions around the North Pacific and indicated that 97.17% of the total diversity was contained within populations and that 2.83% was due to differences between population differences and run-timing differences. Within run timings, 0.80% was due to differences among populations, and 0.27% was due to regional differences. The calculation of diversity within and between run timings was not straightforward, because southern Puget Sound summer-run fish are genetically closer to Puget Sound fall-run fish than to summer-run fish in Hood Canal and the Strait of Juan de Fuca. Summer-run fish in southern Puget Sound were therefore excluded from the calculation of diversity among populations within a run timing, which was 0.91%. The addition of southern Puget Sound fish represented 0.05% of the diversity, and the diversity due to differences among run timings was 0.80%.
To develop a better perspective of regional genetic variability, we combined the allele-frequency data of Phelps et al. (1994) with those of Phelps (footnote 26). The combined data set included allelic frequencies for 34 loci in samples from 116 localities. Multidimensional scaling in three dimensions of Cavalli-Sforza and Edwards (1967) chord distances (Fig. 13) showed four groups of chum salmon: 1) summer-run populations in the Strait of Juan de Fuca (nos. 108-110) and Hood Canal (111-116); 2) fall-, summer-, and winter- run chum salmon in British Columbia (1-26) and Puget Sound (27-74) and fall-run fish in Hood Canal and the Strait of Juan de Fuca (75-95); 3) all samples of chum salmon from outer coastal populations of Washington and Oregon (96-104, 107); and 4) all samples of chum salmon from the Columbia River (105-106).
To depict the genetic relationships among Washington and British Columbia populations, we used multidimensional scaling in two dimensions of chord distances based on 34 loci (Fig. 13). These populations consist largely of fall-run chum salmon, but also of genetically similar winter- and summer-run chum salmon in Puget Sound. Several geographically meaningful clusters appeared in the graph. Samples of fall-run chum salmon from southern Puget Sound (nos. 47-74) and from Hood Canal, or fish of Hood Canal origin (75-89), formed a large cluster. Three samples from Puget Sound summer-run chum salmon (69-72) and from two winter-run populations (73-74) were included in the large Puget Sound cluster of samples. Two overlapping clusters of samples from northern Puget Sound (27-46) and the Fraser River (17-26) were placed next to the large southern Puget Sound-Hood Canal cluster. Clusters of samples from southern British Columbia (12-16), eastern and southern Vancouver Island (4-11), and the Strait of Juan de Fuca (92-95) populations overlapped with each other.
No clear genetic boundaries appeared between adjoining clusters, but more widely separated populations generally showed larger genetic distance than did nearby populations. West Vancouver Island populations (nos. 1-3) were placed in a distinct cluster separate from other British Columbia populations. Samples from Washington and Oregon outer coastal populations (96-104, 107) and Columbia River populations (105-106) formed two distinct clusters, in which the nearest genetic neighbors of the Columbia River populations were among the outer coast populations.
We further analyzed allelic frequencies of the fall-run chum salmon samples reported by Phelps et al. (1994) with spatial autocorrelation to test for isolation by distance among populations. The autocorrelation coefficient, Moran's I (Cliff and Ord 1981), was calculated between samples in 13 50-km distance classes for 19 independent alleles and presented as a correlogram (Fig. 14, Table 12). Significant deviations of Moran's I from zero were detected by standard normal testing procedures following the methods described by Sokal and Oden (1978) and Jumars et al. (1977). This statistic detects positive and negative correlations between sample allelic frequencies within a distance class relative to the average allelic frequencies over all samples.
The results for five alleles--sIDHP-2*86, sIDHP-2*36, sMDHA-1*100, sMDHB-1*100, and mMEP-2*100--were typical of those for 14 other alleles and are shown in Figure 15. High positive autocorrelation appeared between allelic frequencies in samples separated by less than about 250 km. Populations separated by distances greater than 250 km do not, on average, influence one another significantly through migration. The neighborhood size for chum salmon populations in the Pacific Northwest is considerably less than the estimated 600 km reported for Japanese chum salmon by Kijima and Fujio (1982), who estimated this parameter by regressing genetic distances between populations on geographic distances. Larger neighborhood size for Japanese populations could be due to the greater number of egg and fry transfers between hatcheries. Among Pacific Northwest samples, significant negative autocorrelation appeared for pairs of samples separated between 250 and 600+ km, except for significant positive autocorrelation for 5 of the 19 alleles between pairs of samples from populations separated by about 300-450 km. These unexpected positive autocorrelations between widely separated populations may be due to genetic similarity from recolonization by the same ancestral populations after deglaciation (McPhail and Lindsey 1986). Alternatively, they may result from long-distance straying induced by alternate migration around Vancouver Island (Beacham et al. 1987) or translocations of chum salmon eggs and fry (Phelps et al. 1994).
The amount of genetic variability in a population reflects the effects of past population events, such as random genetic drift, the introduction of variability through immigration or mutation, and natural selection. If the effects of mutation and natural selection on electrophoretic variability are assumed to be minimal, reduction in genetic variability within a population may be taken as an indication of past reductions in population size. Larger populations are less subject to the loss of genetic variability through random genetic drift. For example, Kijima and Fujio (1984) found a significant correlation (R = 0.40, P < 0.05) between average, direct-count heterozygosity and population size in 37 river populations of chum salmon in Japan.
In a study of Asian chum salmon populations by Winans et al. (1994), gene diversities (expected heterozygosities) in over 62 loci (including monomorphic loci) ranged from 0.066 to 0.087 among samples, averaging 0.079. However, no regional trends were apparent in the geographical distributions of these values. Overall, this level of genetic variability represents a large amount of genetic diversity relative to that found in other vertebrates (Ward et al. 1992). For northwestern Alaska and Russian populations, Wilmot et al. (1994) estimated expected heterozygosities from 54 loci and found values ranging from 0.056 to 0.072. Heterozygosities averaged by Wilmot et al. (1994) over populations within regions were 0.064 among lower Yukon River summer-run populations, 0.062 among upper Yukon River fall-run populations, 0.065 among Bristol Bay populations, 0.064 among Alaska Peninsula populations, and 0.063 among Russian populations. In 83 chum salmon populations in British Columbia, Beacham et al. (1987) used 9 loci to estimate expected heterozygosities in 5 areas and found that populations in the Fraser River and on the south coast of British Columbia tended to have smaller values. However, population abundances, and presumably effective population sizes, are smaller in northern areas, especially on Queen Charlotte Island, than in southern areas of British Columbia (Beacham 1984). Since Beacham et al. (1987) used only 9 polymorphic loci, their heterozygosity values cannot be compared directly to other studies that used 50 or more loci in their estimates of heterozygosity.
Phelps et al. (1994) estimated population heterozygosity from direct counts of heterozygotes in a set of data that included 38 polymorphic loci and that overlapped with the sets of loci used by Beacham et al. (1987) and Winans et al. (1994). Since monomorphic loci were not included in the estimates of Phelps et al. (1994), heterozygosities between the studies are not directly comparable. Average heterozygosities varied among populations from 0.082 to 0.116 and did not deviate more than 0.005 from heterozygosities expected with random mating. The highest heterozygosities occurred in Hood Canal fall-run chum salmon (0.108 on average) and southern Puget Sound fall- and summer-run populations (0.102 on average). The lowest values were found in the Strait of Juan de Fuca (0.090), Georgia Strait (0.090), and the west coast of Vancouver Island (0.089) fall-run populations. Heterozygosities in hatchery populations were nearly the same as heterozygosities in nearby wild populations. These results suggest that chum salmon hatchery populations in Washington have not experienced recent or historical bottlenecks in population size.
Allelic frequencies for a large number of protein-coding loci indicate that chum salmon populations along the rim of the North Pacific are divided into several regional groups. Russian and northwestern Alaska chum salmon appear to be more closely related to each other than either is to Japanese chum salmon. Frequencies of a mitochondrial DNA variant and mini-satellite DNA variants also show a major difference between Japanese chum salmon populations and all other populations. In northwest Alaska, allozyme frequencies indicate that fall-run chum salmon in the upper Yukon River are distinct from summer-run chum salmon in the lower Yukon River. In the eastern North Pacific, populations of chum salmon from central Alaska to Washington appear to be genetically isolated by distance, so that geographically proximate populations are more closely related to one another than populations separated by large distances.
Two major genetic groups are present in central and southern British Columbia, Washington, and Oregon. One consists of summer-run chum salmon in Hood Canal and the Strait of Juan de Fuca, and a second large group consists of fall-, winter-, and summer-run chum salmon in other areas. The second large group is weakly divided into two groups: 1) coastal populations along the outer coast of Washington and Oregon, including those in the Columbia River, and 2) the remaining populations in British Columbia and Washington (including the Strait of Juan de Fuca populations). Levels of genetic variability within and between populations in several geographic areas are similar, and populations in Washington show levels of genetic subdivision which are typical of those seen between summer- and fall-run populations in other areas and which are typical for populations within run types.
Based on a review of the biology and ecology of chum salmon, the BRT identified four ESUs for the species in the Pacific Northwest (Fig. 16). Genetic data (from protein electrophoresis and DNA markers) were the primary evidence considered for reproductive isolation criterion. This evidence was supplemented by inferences about barriers to migration created by natural geographic features. Data considered important in evaluations of ecological/genetic diversity included distributions, migrational and spawning timing, life history, ichthyogeography, hydrology, and other environmental features of the habitat. In the following summaries, we describe those factors that were valuable in making individual ESU determinations.
Each of the ESUs include multiple spawning populations of chum salmon, and most ESUs also extend over a considerable geographic area. This result is consistent with NMFS species definition policy (Waples 1991:20), which states that in general, "ESUs should correspond to more comprehensive units unless there is clear evidence that evolutionarily important differences exist between smaller population segments." However, considerable diversity in genetic or life-history traits or habitat features may exist within a single complex ESU, and the descriptions below briefly summarize some of the notable types of diversity within each ESU. This diversity is considered in the next section in evaluating risk to the ESU as a whole.
The Puget Sound/Strait of Georgia ESU (Fig. 16) includes most U.S. populations of chum salmon and the vast majority of adult spawners that return to U.S. waters outside Alaska. This ESU includes all chum salmon populations from Puget Sound and the Strait of Juan de Fuca as far west as the Elwha River, with the exception of summer-run populations in Hood Canal and along the eastern Strait of Juan de Fuca. The BRT concluded that this ESU also includes Canadian populations from streams draining into the Strait of Georgia. A northern boundary for this ESU was tentatively identified as Johnstone Strait, but this determination was hampered by a paucity of information from populations in central and northern British Columbia. Chum salmon from the west coast of Vancouver Island are not considered part of this ESU, in part because available genetic information suggests these fish are distinct from Puget Sound or Strait of Georgia fish.
Genetic, ecological, and life-history information were the primary factors used to identify this ESU. Environmental characteristics that may be important to chum salmon (e.g., water temperature, and amount and timing of precipitation) generally show a strong north-south trend, but no important differences were identified between Washington and British Columbia populations. An east-west gradient separating Olympic Peninsula populations from those to the east was considered to be more important for evaluating chum salmon populations.
Chum salmon populations within this ESU exhibit considerable diversity in life-history features. For example, although the majority of populations in this ESU are considered to be fall-run stocks (spawning from October to January), four summer-run (spawning from September to November) and two winter-run (spawning from January to March) stocks are recognized by state and tribal biologists in southern Puget Sound. Summer chum salmon in southern Puget Sound are genetically much more similar to Puget Sound fall chum salmon than they are to other summer-run populations in Hood Canal and the Strait of Juan de Fuca. These data suggest relatively weak isolation between summer- and fall-run chum salmon in southern Puget Sound and/or a relatively recent divergence of the two forms. Reproductive isolation of the Nisqually River and Chambers Creek winter-run populations, which are the only populations in the ESU whose spawning continues past January, may be somewhat stronger.
The Nisqually and Puyallup Rivers are also unique in southern Puget Sound because their headwaters are fed by glaciers on Mount Rainier, giving the rivers different characteristics than other regional river systems. The Nisqually population is also one of the more genetically distinctive chum salmon populations in Puget Sound. However, the genetic differences are not large in an absolute sense, and the majority of the BRT felt that the distinctiveness of the winter-run populations was not sufficient to designate these populations a separate ESU. Rather, the team concluded that these populations, along with the summer-run populations in southern Puget Sound, reflect patterns of diversity within a relatively large and complex ESU.
This ESU includes summer-run chum salmon populations in Hood Canal in Puget Sound and in Discovery and Sequim Bays on the Strait of Juan de Fuca. It may also include summer-run fish in the Dungeness River, but the existence of that run is uncertain. Distinctive life-history and genetic traits were the most important factors in identifying this ESU.
Hood Canal summer-run chum salmon are defined in SASSI (WDF et al. 1993) as fish that spawn from mid-September to mid-October. Fall-run chum salmon are defined as fish that spawn from November through December or January. Run-timing data from as early as 1913 indicated temporal separation between summer and fall chum salmon in Hood Canal. Even though for many years there have been hatchery releases of fall chum salmon in Hood Canal of about 35 million fish annually, and many25 of these fish return to hatcheries in Hood Canal and were historically spawned before the end of October, recent spawning surveys show that temporal separation still exists between summer and fall chum salmon. Genetic data indicate strong and long-standing reproductive isolation between chum salmon in this ESU and other chum salmon populations in the United States and British Columbia. Hood Canal is also geographically separated from other areas of Puget Sound, the Strait of Georgia, and the Pacific Coast.
In general, summer-run chum salmon are most abundant in the northern part of the species' range, where they spawn in the main stems of rivers. Farther south, water temperatures are so high and stream flows are often so low during late summer and early fall that conditions become unfavorable for salmonids. River flows typically do not increase and water temperatures do not decrease until the arrival of fall rains in late October/November. Presumably for these reasons, few summer chum populations are recognized south of northern British Columbia. Ecologically, summer-run chum salmon populations from Washington must return to freshwater and spawn during peak periods of high water temperature, suggesting an adaptation to specialized environmental conditions that allow this life-history strategy to persist in an otherwise inhospitable environment. The BRT concluded, therefore, that these populations contribute substantially to the ecological/genetic diversity of the species as a whole.
Some chum salmon populations in the Puget Sound/Strait of Georgia ESU, which has four recognized summer-run populations and two recognized winter-run populations, also exhibit unusual run timing. However, allozyme data indicate that these populations are genetically closely linked to nearby fall-run populations. Therefore, variation in run timing has presumably evolved more than once in the southern part of the species' range. Genetic data indicate that summer-run populations from Hood Canal and the Strait of Juan de Fuca are part of a much more ancient lineage than summer-run chum salmon in southern Puget Sound.
This ESU includes all natural chum salmon populations from the Pacific coasts of Washington and Oregon, as well as populations in the Strait of Juan de Fuca west of the Elwha River. This ESU is loosely defined at present, and is defined primarily on the basis of life-history and genetic information. Allozyme data show that coastal populations form a coherent group that show consistent differences between other fall-run populations in Washington and British Columbia. Geographically, populations in this ESU are also isolated from most populations in the Puget Sound/Strait of Georgia and Columbia River ESUs.
Ecologically, the western Olympic Peninsula and coastal areas inhabited by chum salmon from this ESU experience a more severe drought in late summer and are far wetter during the winter than areas in the Puget Sound/Strait of Georgia region. All chum salmon populations in this ESU are considered to include fall-run fish. Some Oregon populations are the only known locations to which 2-year-old adult chum salmon return with any appreciable frequency.
Chum salmon from this ESU cover a large and diverse geographic area--from the Strait of Juan de Fuca (lat. 48°20�N) to at least southern Oregon--and the historic ESU may have extended to the recorded extreme limit of the species' distribution near Monterey, California (lat. 36°50�N). Many BRT members concluded that multiple ESUs of chum salmon may exist in this area, but a more detailed evaluation was hampered by a scarcity of biological information of all types. It is possible that many (perhaps most) reports of chum salmon in California and southern Oregon do not represent permanent spawning populations, but rather episodic colonization from northern populations. Even if this is the case, however, the southern limit to permanent natural populations is unclear.
The boundary between this ESU and the Puget Sound/Strait of Georgia ESU is uncertain, particularly with respect to fall chum salmon in the Dungeness and Elwha Rivers. Genetic data for these two populations are ambiguous (Elwha--because of hatchery stocking) or nonexistent (Dungeness), and run timing is also largely uninformative regarding the affinities of these two populations. Although coastal populations generally return and spawn slightly earlier than those in Puget Sound, there is little difference in run timing between Puget Sound and Strait of Juan de Fuca populations. The Washington Department of Fish and Wildlife (Phelps et al. 1995) considers the Dungeness and Elwha River populations to be affiliated with Strait of Juan de Fuca populations to the west, primarily because of their geographic separation from inner Puget Sound fall-run populations. However, the transition to the wetter, coastal climate occurs west of the Elwha and Dungeness Rivers on the Olympic Peninsula. After considerable discussion, the BRT concluded, based on available information, that fall chum salmon from the Dungeness and Elwha Rivers should be considered part of the Puget Sound/Strait of Georgia ESU.
The BRT concluded that, historically, there was at least one ESU of chum salmon in the Columbia River. Ecologically, Columbia River tributaries differ in several respects from most coastal drainages. Genetic data are available only for two small Columbia River populations, which differ substantially from each other as well as from all other samples examined to date.
Historically, chum salmon were abundant in the lower reaches of the Columbia River and may have spawned as far upstream as the Walla Walla River (over 500 km inland). Today only remnant chum salmon populations exist, all in the lower Columbia River. They are few in number, low in abundance, and of uncertain stocking history.
The question of the extent of the Columbia River ESU along the Washington and Oregon coasts prompted considerable debate within the BRT. The BRT concluded, based upon the genetic and ecological data available, that chum salmon in the Columbia River were different enough from other populations in nearby coastal river systems (e.g., Willapa Bay, Grays Harbor, Nehalem River, and Tillamook River) that the Columbia River ESU should extend only to the mouth of the river.
Busack and Shaklee (1995) identified Major Ancestral Lineages (MALs) and Genetic Diversity Units (GDUs are subdivisions of MALs) for several salmon species in Washington. This effort, which sought to identify the existing amount and patterns of genetic diversity within the state, supports the goals of the Wild Salmonid Policy under development by state and tribal fishery managers and is intended to facilitate its implementation. The terminology (GDUs and MALs) differs somewhat from that of previous documents prepared by WDW and WDFW (e.g., Leider et al. 1995). According to Busack and Shaklee (1995), GDU designations were based on a combination of genetic, life-history/ecological, and physiographic/ecoregion data. The authors also stated that they expected that individual ESUs would often include multiple GDUs but would be unlikely to include multiple MALs.
The geographic boundaries of the proposed ESUs for chum salmon are largely consistent with the GDUs and MALs identified by Phelps et al. (1995). With respect to populations in Washington, three of the ESUs for chum salmon proposed here (Puget Sound/Strait of Georgia, Hood Canal Summer-Run, and Coastal) are similar to MALs identified by WDFW. Each of the three MALs identified by WDFW included multiple GDUs. As noted above, one difference between the NMFS and WDFW frameworks is that Phelps et al. (1995) consider all fall chum salmon from the Strait of Juan de Fuca to be in the Coastal MAL, whereas the BRT includes the Dungeness and Elwha River fall-run populations in the Puget Sound/Strait of Georgia ESU. Another difference is that the BRT recognize a Columbia River ESU, whereas Columbia River populations were considered by Phelps et al. (1995) to be part of two GDUs within a larger MAL that included coastal and Strait of Juan de Fuca populations.
As part of an effort to complete comprehensive status reviews for all anadromous Pacific salmonids, NMFS has made ESU determinations for coho and pink salmon and steelhead from the same geographic areas covered by this status review for chum salmon. Although there are similarities in the geographic coverage of ESUs for these species, each species has distinctive attributes that merit special consideration.
The Puget Sound/Strait of Georgia ESU for chum salmon proposed here is similar in geographic coverage to ESUs for coho and odd-year pink salmon. As was the case for chum salmon, genetic and life-history data for both these species showed substantial genetic similarities between U.S. and Canadian populations. Although no recent genetic data are available for comparing U.S. and British Columbia steelhead populations, life-history data showed an abrupt change in smolt age at approximately the U.S.-Canada border, and this was a significant factor in the determination that the Puget Sound ESU for steelhead does not include populations from British Columbia. The western boundary for the Puget Sound/Strait of Georgia chum salmon ESU is similar to the boundaries proposed for coho, pink, and steelhead salmon ESUs.
The Hood Canal Summer-Run and Pacific Coast ESUs for chum salmon have no clear analogues in the other species for which comprehensive status reviews have been completed. However, the even-year pink salmon ESU (which contains a single U.S. population) shares with the Hood Canal Summer-Run ESU the features of restricted geographic range and a small number of component populations. The geographic coverage of the Pacific Coast ESU for chum salmon includes areas that are inhabited by steelhead and coho salmon from multiple ESUs. It is possible that additional information will indicate that multiple ESUs of chum salmon also occur in this area. The geographic extent of the Columbia River ESU for chum salmon is similar, but not exactly congruent with ESUs for coho salmon and steelhead in the same area.